Multi-level programming, as defined in this paper, refers to a problem situation wherein outer decision makers affect the solution space of inner decision makers within a strictly hierarchical structure. The problem corresponds to a primordial economic policy problem since: (i) higher level decision makers have direct control of some variables (the policy variables), but not all variables; (ii) lower level decision makers have direct control of other variables (the behavioural variables), which are manipulated in the light of the levels of the policy variables; and (iii) the higher level decision makers wish to influence a third set of impact variables which are frequently not under the direct control of either decision making group. This problem structure contrasts markedly with mathematical programming, where all variables are under direct control of one decision maker. After a review of the geometry of the multiple-level programming problem, this paper includes a discussion of algorithmic principles, presentation of and comments on an algorithm which allows an implicit search of policies which affect resource availabilities for the behavioural problem, and a numerical example.